Framework · v2.0 · Current
The Second Door Plan
Turn one lesson from your experience into something another person can use.
- Introduced
- September 7, 2026
- Origin
- Episode 1
- Territories
- Work & AI, Uncertainty & Agency, Identity & Worth
The question it addresses
What can I do with the lessons my experience gave me, now that part of the work I learned is changing?
A second door is another place where you can use something you know to help someone with a problem. Choose one problem you understand. Make a small first version. Let someone try it, improve what fails, and share what you learned. You can repeat the process with another problem and gradually build a collection of useful work.
- Notice
- Build
- Test
- Improve
- Share
Start here
People ask me for help when…
Choose one answer. Write the first question you usually ask them.
The framework
The five steps
Each step expands. Work through them in order. Steps 3 and 4 need another person, so expect to pause there.
Not a skill and not a job title. A problem: something that goes wrong in a predictable way, that people already come to you about.
Say who faces it and what happens when it goes wrong. Keep the situation small. "Help teams work better" is too large to try. "Help a team check approvals and availability before promising a delivery date" is small enough.
Ask yourselfPeople ask me for help when…
Give the page a plain name. Write the questions or steps in the order you use them. Say when to use it, and say what it is not for.
Add one filled-in example so a reader can see it working. Keep the first version small, one page if you can. A tool can help shorten the words. Check every suggestion against your own experience, and do not upload anything confidential.
Ask yourselfWhat are the questions you already ask, in the order you ask them?
Ask plainly: "Would you try this on a small task? I want to find the confusing or missing parts." A past example that already went wrong works well, because you both know how it ended.
Watch where they need help and what the method misses. Record what actually happened, separately from anything kind they said about it. A compliment or an interested nod does not establish that the method is useful.
Ask yourselfWho could try this on something small, and what will you watch for?
Change the part that did not work. Where it is practical, let someone outside your usual team try the revision, because they will not fill in the assumptions your colleagues share with you.
Record the next result and the limits. Note what else changed that might explain it. Write down what you can honestly say so far and what you still do not know. If the method still does not help, simplify it, choose a more specific problem, or set it aside.
Ask yourselfWhich part failed, and what will you change about it?
Start with one person or group that has this problem. Public posting is optional. Get permission for anything that identifies another person or their work.
Say what you kept seeing, what you tried first, what actually happened, what you changed and why, and what you have not tested. Then ask where the method might still fail. That question is the part that keeps improving it.
Ask yourselfWho has this problem, and what will you ask them to look for?
Example
Worked through
A fictional teaching example. Maya is invented, not a customer case or an observed success.
Maya has spent fifteen years helping teams finish projects.
A customer asks for a brochure by Friday. The team agrees. On Friday, the printer is still waiting for approved artwork. The person who can approve it has been away all week. The customer needs the brochure for a Monday event.
Notice. Maya recognises the repeating problem: people promise a final date before checking what must happen first.
Build. Version 1 asks four questions.
- What are we promising?
- Who will do the work?
- When must it arrive?
- What happens if it is late?
Test. A colleague tests it by replaying the delivery that already went wrong. Every answer can be filled in: brochure, design team, Friday, Monday event. The colleague points out that the page still does not ask who must approve the artwork. Version 1 would miss the same problem.
Improve. Version 2 adds: "What must happen first? Who needs to help or approve it?"
Someone from an events team tries this version on a separate plan: chairs must arrive before a room can be set up. They name the supplier, identify the approver and write Thursday as the delivery date. Then they ask whether the supplier has actually agreed to Thursday. Maya had assumed someone would make that call.
Version 3 adds: "Has each person confirmed they can do their part by that date?"
They call. Thursday is unavailable. They confirm Friday delivery and Saturday setup before promising the organiser when the room will be ready.
Share. Maya explains both missing questions, her two revisions and the limits, and asks readers where the method might still fail.
The finished page
What Maya ended up with
This story is fictional. Its attempts and results explain the process. They are not evidence of real-world effectiveness.
Worked example: before you promise a date
| Step | What happens |
|---|---|
| Notice | A brochure was promised for Friday. Artwork approval was missing. A Monday event made the delay matter. |
| Build | Maya writes four questions: what is promised, who does the work, when it must arrive, and what happens if it is late. |
| Test | A colleague replays the failed delivery using the page. Every field can be filled in without naming the approver. Version 1 misses the same problem. |
| Improve | Version 2 asks what must happen first and who must help or approve. An events team tries it and spots another assumption: nobody has confirmed the supplier can meet the date. Version 3 adds that check. |
| Share | Maya explains both missing questions, her revisions, and the limits in an article. She invites readers to identify other situations where it fails. |
In the second fictional attempt, the supplier cannot deliver on Thursday. The team confirms Friday delivery and Saturday setup before making a promise to the organiser. The observed result is a changed plan with confirmed timing. We have not shown that delivery actually happened.
Completed version 3: before you promise a date
Use when finishing something depends on other people. This page cannot guarantee delivery or replace any required professional review.
| Question | Events-team example |
|---|---|
| What are we promising? | Chairs delivered and room arranged before the organiser needs it. |
| What must happen first? | Order approval, supplier delivery, room setup. |
| Who must do or approve each part? | Event lead approves. Supplier delivers. Room team sets up. |
| Has each person confirmed their part and date? | Event lead: Wednesday approval. Supplier: Friday delivery. Room team: Saturday setup. All confirmed in this fictional example. |
| What happens if a date changes? | Tell the organiser and review the plan before making a new promise. |
Over time
What can grow from one useful idea?
Read this part after someone has actually tried your method. Until then it describes a road you have not walked yet.
Your experience has now been used somewhere beyond your usual job. And you can do this again. The next time you notice a problem you know how to solve, go through the same process. Make something someone can use. Let them try it. Learn what needs to change.
Over time you might have five useful ideas. Perhaps fifteen. Each with an example, a way to use it, and something you learned from trying it. Now you have more to share.
Where a collection can go
You could bring those ideas to a professional group. Put them on a website. Teach a short workshop. Perhaps some of them belong together in a book. Start with the place where people already care about the problems you are solving.
Someone might read one of your examples and ask you to explain it to their team. That conversation could lead to another invitation. Perhaps, eventually, a conference.
It could happen inside your company too. A colleague uses your method. Their manager asks you to share it with the department. People begin coming to you when they face a similar problem. They know something specific about you: you can take a problem apart and give them a useful way to approach it.
What is promised, and what is not
Could that lead to a different role, a raise, or a promotion? Possibly. None of those outcomes is promised, and none of them may arrive.
But look at what you have already built. You have work you can show. You have learned which ideas help people and which still need work. You have somewhere to develop the next one. And you have another way to describe yourself.
- "I help people plan projects without missing the steps that delay them."
- "I explain things people struggle to understand."
- "I make tools that help people check their work."
Your job title can change. You can keep developing that part of yourself.
AI may change the work you spent fifteen years learning. Those years also gave you problems you recognise, questions you know to ask, and lessons you can test somewhere new.
Now you have begun doing that. That is your second door.
Try it here
Walk through it now
The worksheet, one step at a time. Answers stay in this browser only and are never sent anywhere. Print or copy a summary when you finish.
Five steps. Write the first line today, then come back when someone has tried your method.
Answers stay in this browser only. Nothing is sent anywhere, and nothing is stored if you do not type.
Educational reflection material. Not psychological, career, financial, legal, or medical advice, and not a validated instrument. No particular result is promised. Read the full scope statement.
What it helps with
- Turning one lesson you already use into something another person can pick up and try.
- Finding out whether a method helps someone other than you, rather than assuming it does.
- Keeping an honest record of what you tried, what failed, and what changed.
- Building up, over time, a small collection of useful work you can show.
What it does not establish
- It is a proposed practice, not a validated instrument. It has not been tested in a controlled study.
- It does not promise employment, income, an audience, or protection from automation.
- A compliment or an interested nod does not establish usefulness. A first draft is not proof that the method works.
- Finishing the written prompts does not mean the real-world attempts are complete. The attempts are the part that matters.
- A small trial shows what happened in one setting. It cannot establish that your method caused the result.
- It is a reflection and action tool for general educational use. It is not advice for your specific situation.
Practice
How to use it
- 01You can begin with one sentence today. Testing and improving the method will take real attempts over time.
- 02Work through the steps in order. Each one depends on the one before it.
- 03Write things down. A method that stays in your head cannot be handed to anyone.
- 04Steps 3 and 4 need another person, so expect to pause there and come back. That pause is the method working, not a delay.
- 05When you have a result, go back to the step the result points at. Usually that is Build.
Download
The worksheet
Worksheet v2.0
The Second Door Plan Worksheet
To turn one lesson from your experience into something another person can try. Start small. You can complete the first line today. The full process takes real attempts and time to learn from them.
PDF and Markdown. Version 2.0. The full worksheet is also on this page, free to use without downloading.
Educational reflection material. Not psychological, career, financial, legal, or medical advice, and not a validated instrument. No particular result is promised. Read the full scope statement.
Provenance
Where this came from
Research that influenced part of the method, design choices, what is original, and what remains unvalidated. Kept separate on purpose.
Influenced by
- Field evidence that AI assistance narrows the visible output gap between novices and experts is part of why this method starts from a problem you have helped with rather than from an output you can produce. The research describes the conditions. It does not test this method. See the linked sources.
- Polanyi's account of tacit knowledge informed the instruction to write the method down, on the grounds that a practitioner usually knows more than they have said out loud, and that saying it is work worth doing.
Design choices
- The five-step sequence, the names of the steps, and the ordering are Sougata's design choices.
- Starting from a problem people already ask you about, rather than from a skill, is a design choice intended to keep the first version small enough to try.
- Requiring that another person tries the method before it counts as tested is a design choice, not a research finding.
- Suggestions in the worksheet, such as asking someone outside your usual team, are practical defaults.
Original here
- The framing of a second door as another place to use something you know is original to Voyagers Beyond.
- The worked example, including its failed first attempt, is written for teaching and is fictional.
Not yet validated
- Whether following this method changes any career, income, or audience outcome has not been studied.
- Whether the Test and Improve steps reliably surface the missing part of a method has not been measured. The site has no record of documented use beyond its own worked example.
- The five-step sequence is a proposed practice. It has not been compared against any other way of doing this.
Working paper · 2023-04
Generative AI at Work
Erik Brynjolfsson, Danielle Li, Lindsey Raymond · National Bureau of Economic Research · NBER Working Paper 31161; later published in The Quarterly Journal of Economics
Evidence
What the source supports
- In this setting, a generative AI assistant increased average productivity, measured as issues resolved per hour.
- The gains were concentrated among novice and lower-skilled agents. Experienced, high-skill agents saw minimal improvement.
- The authors interpret the tool as spreading the practices of more able workers to newer ones.
Boundary
What it does not establish
- That experienced workers in other fields will see the same pattern. The study covers one company and one kind of work.
- That AI assistance makes experience worthless. The tool learned from the experienced agents' own behavior.
- Any claim about job loss, wages, or long-term careers. The study measures task performance, not employment outcomes.
Interpretation
What Sougata is doing with it
Sougata uses this as context for why Episode 1 exists: in this setting the visible output gap between newer and more experienced workers narrowed quickly. That is the situation the Second Door Plan responds to. The paper does not evaluate the Second Door Plan, and nothing in it supports a claim that writing a lesson down and testing it produces any career outcome. The connection between the two is Sougata's reasoning, not the paper's finding.
Working paper · 2023-09
Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and Quality
Fabrizio Dell'Acqua, Edward McFowland III, Ethan Mollick, Hila Lifshitz-Assaf, Katherine Kellogg, Saran Rajendran, Lisa Krayer, François Candelon, Karim Lakhani · Harvard Business School · HBS Working Paper 24-013; forthcoming in Organization Science
Evidence
What the source supports
- AI capability is uneven across tasks that look similar in difficulty. The authors call this the jagged frontier.
- For tasks inside the frontier, consultants with the tool performed measurably better on speed and rated quality.
- For a task outside the frontier, tool use was associated with worse accuracy.
Boundary
What it does not establish
- That the frontier is fixed. The authors describe it as moving and hard to see from the inside.
- That professionals can reliably tell in advance which side of the frontier a task sits on.
- Any general claim about consulting careers, hiring, or the value of experience over time.
Interpretation
What Sougata is doing with it
Sougata uses this for one specific point: a tool's competence is uneven in ways that are hard to see from inside a task, so knowing where it fails is worth something. This is context about AI and work. It is not evidence that the Second Door Plan helps anyone, and the paper makes no claim about that method.
Book · 1966
The Tacit Dimension
Michael Polanyi · University of Chicago Press (originally Doubleday, 1966)
Evidence
What the source supports
- The claim that skilled practice includes knowledge the practitioner cannot fully articulate.
- The observation that explicit rules never capture the whole of a skill.
Boundary
What it does not establish
- Any claim about artificial intelligence, which the book predates by decades in its modern form.
- The claim that tacit knowledge cannot be imitated in output. Polanyi is describing knowing, not the reproducibility of results.
Interpretation
What Sougata is doing with it
Sougata uses Polanyi for one reason: if a practitioner usually knows more than they have said out loud, then writing a method down is how you find out which parts you had never actually stated. That is why the Second Door Plan has a Build step. Polanyi says nothing about AI and nothing about this method, and the book is not evidence that human judgment is safe from anything. Applying his vocabulary here is Sougata's extension and should be read as interpretation.
History
Version history
Earlier versions are not overwritten invisibly.
- v2.0New five-step sequence: Notice, Build, Test, Improve, Share. The steps now require another person to try the method and record what happened. Replaced the worked example with a fictional story that shows a failed first attempt and a second assumption found by an outside reader. Added a section on what a collection of useful work can grow into. Removed the ninety-minute completion promise. Not validated.
- v1.0First public version, introduced alongside Episode 1. Stages were Find, Verify, Extract, Demonstrate, Test, with a worked example about technical proposals.
Continue the question
Related questions
Work & AI
What were my fifteen years for?
A tool produced in minutes something that resembles work you spent years learning to do. The question is not rhetorical. The useful answer is not a defence of those years. It is a way to find out which lessons in them are still worth something to somebody else.
Uncertainty & Agency
How do I adapt when the target keeps moving?
Every piece of advice says adapt. Almost none of it says to what. When the thing you would adapt to changes every few months, adaptation as a strategy needs a different definition.
Continue